Re: Potentially Predatory Pre-Announcement of Possible Vaporware (zinq-djbdns-0.01)
Laurent Bercot <[email protected]>
| Newsgroups | gmane.network.djbdns |
|---|---|
| Message-ID | <[email protected]> |
>> Two sentences: "Ewww! Why?" > One sentence: Why Not? The burden of proof is always on the side of the one proposing a change. You took a Makefile that worked and was easy to modify, and turned it into an automatically generated configure script, which will automatically generate a Makefile, adding two or three layers of error-prone shell code, for a benefit that's not obvious at first sight. So it's legitimate to ask you "Why?" You may have very good reasons to do what you did, but "why not?" is not one of them. ;) > I don't see how it adds a dependency. At least not to the end user. Let's see. Autoconf-created scripts are usually full of bashisms. So it adds the following requirement: your /bin/sh must be damn close to /bin/bash. Autoconf-created scripts use "head -1 foo" instead of "head -n 1 foo", so it adds a dependency to coreutils that understand the -1 form (which is not required by Single Unix). I could never tell autoconf-created configure scripts that my system was i386-linux with gcc *but* the diet libc, and not the glibc. autoconf makes it harder, not easier, for users of non-mainstream software (when it's supposed to help with portability in the first place). I could say a lot more, but you get the point. DJB didn't use autoconf, and went so far as to write his own build system, for a reason. You're not DJB, and that's fine, but if you're going to take the plunge and package his software, you might as well have a deeper look at his philosophy. Porting things to autoconf has benefits, but it also has costs; this is not a decision to take lightly, and I, among others, am not sure you were aware of all the costs before making the decision. It's your software and you can do whatever you want with it. But who's going to use it if you don't even satisfy the hardcore djbists? :P -- Laurent